[Graeme Nimmo]

I’ve been playing with Stable Diffusion for about a week and a half now and I’m quite happy with my progress so far. I’m getting better at building decent prompts, selecting models and lora to help me take the idea of an image I’ve got from my head and bring it to my screen (I’ve even then printed off a couple of AI images as photographs on an old Photo printer I found at the back of a drawer whilst I was cleaning up the other day) and my latest achievement was training some Loras for SDXL based on Andrew’s awesome guide on here.

But, I notice that in all of the content I see here and more widely online almost all of the content I’m seeing is making reference to Automatic1111 or ComfyUI for running SD on your own local hardware. The first local SD UI I actually got running was Invoke-AI (https://github.com/invoke-ai/InvokeAI) and I’ve been finding it to be immensely powerful, has a good community and when I get a stronger level of base knowledge I’m hoping to leverage some of that power. After becoming a member here I’ve tried Automatic1111 and ComfyUI, but I just kept going back to Invoke. The thing that’s sadly missing from it though is Adetailer, but I’ve seen that it’s possible to implement it using their workflow tools and there’s even a sample one for SD1.5 but not for SDXL (I’ll soon be going to the community for them to see if anyone can provide me with one as they’ve got a section of their community for sharing workflows)

Does anyone else use it and also, if you have used it did you stick with it or go to another tool?

I’d love to see some guides written here for it, but I appreciate that writing guides for two platforms is hard enough without needing to add a third one into the mix!

[Andrew]

I have only used Invoke AI briefly in its early stages. I just took a look, and it seems to have evolved quite a bit! will be interesting to try it out.

[Graeme Nimmo]

Yea, they’ve also got a YouTube channel where they do live demos of things and wow, that universal canvas feature they’ve got is amazing! A recent video showed them making some monsters using a mixture of controlnets, image to image stuff and manual tweaks to the images.

[Author]

Posts